Application of Ant Colony Optimization on GPON ring network design

Abstract

This thesis addresses the problem of optimizing the design of Gigabit Passive Optical Network (GPON) rings. The objective is to determine an optimal configuration that minimizes the total design cost while considering the specific requirements of GPON networks. The traditional approach of manually planning the GPON Ring Design proves to be inadequate. The network planner designs the network while considering all constraints and staying within the budget. This approach falls short in delivering optimal solutions due to its reliance on the expertise and subjective decision-making of network planners. Therefore, an enhanced ACO Algorithm is proposed, which addresses this limitation by proposing an automated and optimized approach using advanced algorithms that incorporates customer proximity-based heuristic information to improve the efficiency and effectiveness of the optimization process.

The research focuses on two main approaches: the TSP-based ACO Algorithm and the Enhanced ACO Algorithm. In the first approach, the ACO Algorithm is built based on the parameters commonly used for TSP problems, where the heuristic information is determined based on the proximity of nodes to one another. The latter approach modifies what determines the nodes’ attractiveness in the heuristic information of the ACO algorithm, favoring nodes that are in closer proximity to customers to be served by the ring. Through this alteration, the ants in the system gain improved insight into what constitutes a good GPON design, specifically in terms of making effective decisions regarding the ring design versus the last-mile cable that reaches the customers. It ensures that the algorithm is specifically tailored to the GPON ring design problem, where the objective differs from that of the TSP-based ACO. The enhancement of the heuristic parameter in the algorithm aligns with the objective of finding near-optimal GPON designs with the goal of maximizing customer proximity.

To enhance the solution stability, the proposed attractiveness parameter is further improved, and its influence is propagated throughout the network. Fine-tuning of the ACO controlling parameters, such as (α, β, and ρ), is carried out to identify the best empirical values. Additionally, the impact of increasing the number of ants in the system is evaluated, considering both solution quality and execution time.

The findings demonstrate that the Enhanced ACO Algorithm outperforms the traditional TSPbased ACO Algorithm in GPON ring design optimization. The Enhanced ACO Algorithm consistently produces more efficient network designs for GPON Ring. Moreover, by comparing the algorithm to an existing MILP, the Enhance ACO Model exhibits fast processing times, offering computational efficiency and quicker solution generation. The algorithm's scalability and reliability are also proven, as it maintains stable performance across networks of varying sizes.

This thesis contributes to the field of network optimization by proposing a tailored algorithm for GPON ring design. The results showcase the effectiveness and suitability of the Enhanced ACO Algorithm, providing valuable insights for practical implementation in GPON networks. The research findings serve as a basis for further advancements in the field of GPON network optimization and can support decision-making processes in the telecommunications industry.

Date of Award	Aug 2023
Original language	American English
Supervisor	Andrei Sleptchenko (Supervisor)